A Configurable Floating-Point Fused Multiply-Add Design with Mixed Precision for AI Accelerators | Publicación